Документ взят из кэша поисковой машины. Адрес
оригинального документа
: http://www.naic.edu/~phil/usrproj/a1552/a1552.html
Дата изменения: Tue Nov 23 01:18:05 2004
Дата индексирования: Sat Dec 22 05:00:31 2007
Кодировка:
Поисковые слова: п п п п п п п п п п р п р п р п р п р п р п р п р п
|
a1552 cordrift mapping lbw, galaxies
nov04
Intro:
A1552 made lband maps using cordrift (drift scans).
The time to drift through the lbw beam was about 15 seconds. To increase
the sensitivity up to 21 separate drift maps were done of the same region.
One map had a relatively strong galaxy present. When the strip that contained
the galaxy was bandpass corrected, a negative image of the galaxy was present
for the entire strip.
The processing steps used for each individual map
were:
-
input a single map using corinpscan
-
Do a linear fit to the cal deflections. Use this rather than the actual
cal deflections. This data was taken when the cal deflection was taken
at the end of the drift scan rather than the beginning (I think..). The
telescope was stationary and the sky was drifting while the cal was turned
on/off. If a continuum source drifted through the beam during this time,
the cal deflection would be contaminated. The linear fit (throwing away
outliers) would be an attempt at fixing this problem.
-
scale the data to kelvins using the cal fits.
-
Do the band pass correction with cormapbc(). Various methods were tried
for the bandpass correction:
-
do a linear fit by channel along the strip direction throwing out outliers.
Use the mean value of the fit for along each channel as the bandpass value
for that channel.
-
compute the median by channel over the strip and use that for the
bandpass correction.
-
Remove Tsys (also in the routine cormapbc()). For each spectra tsys was
computed by taking the median over the channels, or a robust average (throwing
out outliers). This was done after the bandpass correction. The computation
did not use 10% of the channels on each edge of the bandpass.
-
The map was then converted to Jy use the gain curve.
-
This process was then repeated for the 21 different maps.
Strip number 16 (counting from 0) contained the galaxy.
The first image shows the dynamic
spectra of this strip using the median for bandpass correction and
robust average for tsys removal.
-
Top image: This a dynamic spectrum of strip 16 for a single map. The data
has been bandpass corrected using the median bandpass for the strip.
The horizontal stripes are coming from a continuum source (720 mJy max,
about 500 mJy this strip). The upper/lower strips are the sidelobes. The
vertical strip on the right is from the galaxy. (note the horizontal axis
is labeled freq Mhz but it is actually channel number ...).
Bottom image: This is the sum of 21 maps of strip16. A median bandpass
over each strip was used and then a robust average for each spectra removed
Tsys. After conversion to Jy, the 21 maps where then added together. There
is a shadow of the continuum source present. The galaxy image goes through
then entire strip.
The next plot shows the total
power along strip 16 computed over the frequency channels that contain
the galaxy (.ps) (.pdf)
(408.5 to 410 Mhz).
-
The units are Tsys. The continuum source peaks at strip sample 67. The
galaxy appears in samples 95 through 115 (bracketed by the dashed lines).
The plot shows why the median bandpass is not going
to work. The plot shows that over have of the samples have either source,
sidelobe, or galaxy. So the median is not going to do a good job. The negative
image of the galaxy in the "other" samples is caused by the bandpass correction
containing some of the power from the galaxy. A robust average was also
used but it did not do any better at removing the galaxy from the bandpass
correction. This is probably because the noise in a single channel is large
enough to not exclude the galaxy samples.
The problem is that there are not enough uncontaminated
samples to compute a good bandpass correction. You could take more samples
(a longer strip) or remove the continuum sources before computing the bandpass
correction. This removal is a bit tricky (especially if there is absorption
in the continuum source spectra).
To solve this problem mary putman suggested breaking
the strip of 120 samples up into smaller sections, computing the median
of each of these sections, and then taking the minimum value of these sections
for the bandpass correction.
The next image shows the band
pass correction breaking the strip up into 3 sections and computing
the median on each section.
-
Top image. This is the same single map image with just a median bandpass
as the previous image.
-
Bottom image. This is the sum of 21 maps using the 3 section median as
the band pass correction. The continuum has been removed similar to the
first image.
Breaking the strip up into 3 sections does not leave a negative image of
the galaxy in the "other" samples. There is still a trace of the continuum
source in the image. The bandpass correction does not contain any of the
continuum source. After the bandpass correction, a single number per spectra
is used to remove the system temperature. Any slope in the continuum flux
would not be removed by the bandpass correction. The average over the resulting
spectra would then leave a slope. Looking at the amount of the slope, the
spectral index would have to be a little larger than .75 to explain all
of the slope.
processing:usr/a1552/new/pg1211/do_3.pro
home_~phil